Skip to content

Go live · Stage B — deploy to the UpCloud cluster #7

Description

@woksin

Stage B of the go-live playbook: move Prompter onto the UpCloud cluster that runs Studio. The Pulumi stack is written and builds, but has never been applied.

About an hour, plus DNS propagation.

Gate: finish #6 first, including the A6 calibration. Every problem found on a laptop is one not debugged through kubectl logs, and an uncalibrated threshold should not meet the community.

  • B1 PULUMI_CONFIG_PASSPHRASE and UPCLOUD_TOKEN repository secrets; confirm the [self-hosted, linux, cratis] runner is available to this repo
  • B2 clusterId and ingressHost in Deployment/Pulumi.production.yaml
  • B3 DNS record pointing at the cluster's ingress load balancer — before B5
  • B4 Runtime secrets via Deployment/scripts/set-secrets.sh, then pulumi stack init production
  • B5 pulumi preview, pulumi up, commit the state
  • B6 First production index through POST /reindex
  • B7 Documentation repo calls /reindex after its deploy
  • B8 Bot installed on the real Cratis server, channel ids in the stack config
  • B9 Privacy notice pinned (P-23)

Two things that fail quietly: the self-hosted runner label queues forever rather than erroring if this repository cannot see it, and a production environment with required reviewers pauses the deploy. Neither should be discovered during a release.

The first pulumi up is where certificate issuance and volume binding usually stumble — read the preview before applying it.

Then: a week of answering only before #8's C2/C3.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions